FCT: DEE - Dissertações de Mestrado
URI permanente para esta coleção:
Navegar
Entradas recentes
- A 1ps time-to-digital converter in CMOS technologyPublication . Rodrigues, Luís Martins; Oliveira, LuísThis thesis presents the design and implementation of a 1ps Time-to-Digital Converter (TDC) in CMOS technology, focusing primarily on a vernier TDC architecture capable of achieving this resolution. A classical delay line TDC was also implemented as a reference to evaluate the effectiveness of robustness-improvement methods. A central objective of this work is to enhance performance stability against process, voltage, and temperature (PVT) variations, assessed through extensive Monte Carlo and corner simulations. The study begins with baseline TDC implementations without robustness mechanisms, followed by versions incorporating a biasing circuit, and finally an enhanced design featuring a Low-Dropout Regulator (LDO) and Bandgap reference circuit. Comparative analyses highlight the impact of these techniques, demonstrating their effectiveness in producing a high-resolution, PVT-robust vernier TDC in modern CMOS processes.
- Performance Evaluation and Lifespan Analysis of Photovoltaic Systems in Desert Regions Compared to Temperate RegionsPublication . Amiry, Sajida; Pereira, PedroThis thesis evaluates the performance and lifespan of photovoltaic (PV) systems in desert regions compared to temperate climates, focusing on the Western Desert of Egypt and Schleswig-Holstein, Germany. The research aims to quantify how solar irradiance, ambient temperature, humidity, and dust affect PV efficiency and long-term reliability. A multi-source methodology was adopted. Meteorological data on irradiance and temperature were retrieved from PVGIS, while humidity data were obtained from the NASA POWER database and dust concentration values from the NASA Giovanni platform. Empirical degradation coefficients (α for dust and β for humidity) were incorporated from published field studies. These datasets were combined to calculate monthly reductions in PV performance. In addition, PVsyst software was employed to model annual energy yield, performance ratio (PR), and system losses for a 1.117 MWp installation at both sites. The results show that while Egypt receives much higher solar irradiance (930.8 W/m² vs. 233.9 W/m² in Germany), the extreme desert climate significantly reduces system efficiency. In the Sahara, high cell temperatures (65.39 °C) lowered module efficiency to 20%, while dust and humidity caused additional performance losses of ~10%. In Germany, the cooler climate (cell temperature 28.6 °C) sustained 23% efficiency, with only ~6% losses, giving temperate conditions an overall 7.21% performance advantage. Simulation results further highlighted this trade-off. In Egypt, the PV system generated 1406.1 MWh/year energy with a PR of ~0.699, primarily constrained by soiling (−20%) and thermal losses (−4.4%). In Germany, the same system produced 562.2 MWh/year energy with a much higher PR of ~0.881, reflecting clean, cool conditions with negligible soiling. Lifespan analysis showed degradation rates of 1.22–2.73% annually in desert climates, limiting modules to 5–10 years of operation, versus ~25 years in temperate regions. The findings confirm that desert regions offer unmatched energy potential but face severe durability challenges. Overcoming these issues requires advanced cooling, robotic or electrostatic cleaning, and dust-resistant coatings. With such technologies and supportive policies, desert solar farms can become a cornerstone of the global renewable energy transition, ensuring both high yield and sustainable reliability.
- Enhancing Information Assurance in Higher Education Institutions: A Technological ApproachPublication . Zorro, Tiago Manuel Morais; Sarraipa, João; Silva, JoãoThe certification process within Higher Education Institutions is important for ensuring academic integrity and authenticity in credential verification. This thesis explores the integration of Physically Unclonable Functions with blockchain technology as a mean to provide a cost-effective and secure authentication mechanism for HEI certificate validation. The primary objective is to design, implement, and evaluate a component authentication system that leverages PUFs as a trust anchor within a blockchain network shared among HEIs and stakeholders. A comprehensive analysis of Information Assurance technologies, Distributed Ledger Technology, Blockchain, and Physically Unclonable Functions is conducted to establish a theoretical foundation for the proposed system. The component is developed to function as an authentication interface between a blockchain network and stakeholders, ensuring the immutability and integrity of published academic records. This component was implemented in a project were there was a need to enhance information assurance within an academic environment. A comparative analysis was then performed between the developed module and alternative solutions. The results indicated that the proposed system effectively authenticates users while ensuring certificate integrity with low-budget resources. The research findings demonstrate that integrating Physically Unclonable Functions with blockchain offers a robust solution for securing academic certification processes. Physically Unclonable Functions provide strong, tamper-resistant authentication through their unique and unpredictable characteristics, while blockchain ensures the integrity and decentralization of the certification data, making the entire process more secure and reliable.
- Carregamento Inteligente de Veículos Elétricos: Otimização de Padrões de CarregamentoPublication . Lopes, João Miguel Travassos Banha; Amaro, Nuno; Reis, FranciscoO exponencial crescimento dos veículos elétricos tem pressionado as redes de baixa ten- são, uma vez que o carregamento simultâneo de múltiplos veículos gera um aumento significativo nos esforços da rede, em resultado da maior procura energética, originando variações na qualidade das tensões da rede e existência de sobrecargas nos barramentos. Este trabalho tem como objetivo desenvolver metodologias capazes de aumentar a capacidade de alocação de pontos de carregamento sem recorrer a reforços imediatos na rede. A abordagem proposta, implementada em Python, assenta numa metodologia de duas etapas: (i) dimensionamento (sizing) e localização (siting) otimizada de cada ponto de carregamento; (ii) operação de veículos elétricos num horizonte de 24 horas, comparando dois modos de carregamento distintos (não otimizado e otimizado). Estas metodologias foram implementadas em P6+ython através da biblioteca Panda Power. Estas metodologias foram aplicadas na rede CIGRE LV, tendo demonstrado que a loca- lização tem tanto impacto como a dimensão da potência instalada, sendo igualmente importante quando comparados ao modo de carregamento selecionado. Em operação o smart charge nivelou os picos de carga, melhorando o perfil das tensões e reduzindo as perdas associadas, mitigando assim a necessidade de reforços imediatos na rede.
- Design of the 3rd Generation Reconfigurable IoT Node with AI-Enhanced Edge ComputingPublication . Julião, André Filipe Lopes; Oliveira, JoãoThe growth of Internet of Things (IoT) has lead a digital transformation, connecting billions of devices that are able to sense, collect, process and exchange information across multiple domains, ranging from everyday applications, to industrial systems. As IoT networks continue growing, integrating edge computing Artificial Intelligence (AI) technology has become a necessary enhancement in intelligent data processing. By enabling a device to analyze and act on data locally, opportunities for autonomous and faster decisions arise, instead of relying on cloud infrastructure. Conventional IoT nodes face critical challenges, as relying on cloud computation introduces issues regarding latency, energy consumption and data privacy. Additionally, with countless devices constantly acquiring and transmitting data, network saturation issues can lead to a system’s impairment. By employing reconfigurable, low-power techniques in combination with AI processing, IoT technologies can move towards an increase in their own autonomy and self-adaptability to any environment. This dissertation will focus on the research and implementation of efficient hardware and firmware that enable local data acquisition, offline preprocessing and intelligent infer- ence, whilst simultaneously maintaining a low-power consumption profile. In addition, a distributed architecture network was designed to demonstrate how edge computing can enhance IoT deployments.
- Techno-Economic Assessment of a Superconducting Fault Current Limiter for Wind Farm Grid Integration: A Case Study in PortugalPublication . Vicente, João Afonso das Neves Santos Bicho; Pina, João; Amaro, NunoThe integration of new renewable energy sources, particularly wind farms, into existing power grids presents significant technical challenges due to the rise in short-circuit current levels. As fault currents approach equipment ratings, grid operators must either reinforce protection systems or consider alternative solutions to ensure safe and reliable opera- tion. Conventional approaches, such as upgrading switchgear and circuit breakers, often involve significant investment and prolonged downtime. High Temperature Supercon- ductor (HTS) Fault Current Limiters (FCLs) of the saturated-core type offer a promising alternative by reducing fault current levels, thereby limiting the need for costly protection system upgrades or allowing such investments to be deferred. Although HTS-based Superconducting Fault Current Limiters (SFCLs) have been successfully demonstrated, they have not been adopted by utilities, primarily due to high initial investment costs. In this context, the present study conducts a Techno-Economic Assessment (TEA) of the Saturated Iron Cores Superconducting Fault Current Limiter (SIC-SFCL) to evaluate its feasibility in practical power system environments. A case study in Portugal is presented, analyzing the impact of the integration of a new wind farm into the Portuguese distribution grid, with the objective of determining whether the implementation of a SIC-SFCL is more economically advantageous than upgrading conventional protection equipment (switchgear and circuit breakers). The economic analysis identifies the conditions under which the SFCL deployment be- comes cost-effective, offering valuable insights for utilities and grid operators considering the integration of superconducting technologies. This assessment therefore contributes to overcoming barriers to SFCL deployment and supports the transition towards more resilient and renewable-based power systems.
- ASSESSMENT OF VEGETATION CHANGES BASED ON REMOTE SENSING IMAGES. A Threshold-Based Time-Series Analysis for Detecting Vegetation CutsPublication . Sacavém, Matilde Campos Coelho Pinto; Fonseca, JoséThis thesis presents a methodology for detecting vegetation cuts using multi-temporal remote sensing data within the Primary Network Fuel Break Lines (PNFBL) and Urban Area Fuel Management Zones (UAFMZ) in northern Portugal. Detecting vegetation cuts consistent with maintenance operations is essential fuel management and wilfire prevention. The study utilized Sentinel-1 and Sentinel-2 imagery from 2022 to 2024 across three designated study areas in northern Portugal, along with a reference dataset developed specifically for performance validation. Two independent pre-processing pipelines were implemented to address the distinct characteristics of active (Sentinel-1) and passive (Sentinel- 2) remote sensing data. These pipelines produced vegetation index images that served as input for a pixel-based time-series analysis aimed at identifying abrupt declines associated with vegetation cutting events. Several methodological components, including image co- registration, temporal compositing, and post-processing steps, were systematically evaluated to assess their impact on detection performance. EVI and NDVI emerged as the most reliable indices, while RVI was found to be too noisy for consistent detection. Optimal threshold settings (decline = 0.7, recovery = 0.5), combined with co-registration and post-processing, achieved the best balance between recall and precision, with an F2-score of up to 0.46, recall of 0.63, and precision of 0.22. Detection performance varied across spatial landscapes and temporal seasons.
- Robotic manipulation of soft-tissues in agri-food processingPublication . Fino, Diogo Lúcio; Rocha, André; Freitas, Nelson; Arvana, MiguelMany challenges in the food industry have arisen from the rapidly growing population increasing the food demand and, by consequence, creating a need to produce higher quan- tities of food products. Additionally, the growing concerns about climate change and the increased environmental awareness among the population have created requirements for the industry in the food sector to adopt more efficient practices with lower environmental impacts. To combat these problems, many businesses have deployed robotic systems, which have already proven to be an effective way to boost productivity rates and reduce the creation of waste. In the food industry there is also the concern with hygiene and po- tencial contamination problems, these concerns create some challenges for the employed robotic systems. The proposed work for this dissertation is the creation of a robotic system to manipulate food products, specially smoked and cured meats like chouriço. The system is composed of two parts, one dedicated to the manipulation and the other dedicated to the vision capabilities of the system in order to identify and locate the various food items. The manipulation part of the system is composed of fin ray effect gripper, 3D printed using TPU filament. This type of gripper is capable of deforming around object creating a more secure grasp on the various products, while limiting the risk of damage to them. The developed object pose estimation algorithm, for the vision part of the system, makes use of certain keypoints on the products to calculate their pose in relation to the camera system. This approach does not require complex data about the objects to be detected, like point clouds. It just the requires a general knowledge about the shape of the products. The developed fin ray effect gripper proved to be adequate in handling the three chosen test products. The developed object pose estimation algorithm archived a success rate of around 80% for products that lay flat on the workspace table. Around the same success rate for the products that present an angle against the table. And a lower rate, of around 72% in cases of overlapping products.
- Implementation of an Intelligent Virtual Assistant based on LLM for Irrigation OptimizationPublication . Chia, Henrique Duarte Mota dos Santos; Oliveira, AnaNew AI assistants are surging, and the technology is already being integrated into day-to-day life and even across organizations. The agriculture field is no exception, and this is a great opportunity to make knowledge available to a wider range of stakeholders across the sector. This technology can prove useful as it can help with information such as the ideal times of planting and the predicted time of plant growth, the protocol treatments for each plant to prevent diseases and optimize growth. This thesis explores the implementation of an AI assistant with the limitations of the available computing resources and strict information security, focusing on the development of an AI assistant that can help the user in all the steps to cultivate a healthy and productive crop yield. The system will be integrated into the already existing platform, Irristrat, which is a monitoring system for the crops, where many factors are shown, such as soil humidity, temperature, … The main factor that led the work carried out was the vision that the assistant developed would use only controlled information and documents where the information is viable and trustworthy. The thesis was carried out in collaboration with Hidrosoph, the organization that developed the Irristrat platform, and installs the sensors used with it. Hidrosoph contributed to the supervision of the work and provided support with the implementation and validation of the system developed in the scope of the thesis. With the limitations of resources available, the solutions discussed always had a factor in common, which is the use of an already existing Large Language Model. Some other options were discussed, like the fine-tuning method, but the RAG method is what ended up being explored. The system proposed in this thesis involves the integration of a vector database, with the role of saving all the information fed to the system, with the main objective of being easy to access/gather. This allows for the system to be possible to be used in a real-time scenario, together with a locally run Large Language Model (LLM) in charge of understanding the user's questions/conversation and using the gathered information to formulate a meaningful answer for the user After the implementation, the system was evaluated in terms of response quality and performance. The results show that the proposed RAG-based assistant is capable of generating coherent and context-aware agronomic responses, demonstrating acceptable performance under constrained computational resources. A qualitive evaluation conducted by agronomy engineers confirmed the relevance and technical consistency of the generated answers, highlighting the feasibility of the proposed approach as a decision-support tool for irrigation optimization.
- AI-Driven Automation and Interoperability for Digital Twins: Leveraging AAS and Generative AIPublication . Troeira, João Carlos Zacarias; Maló, Pedro; Marques, FranciscoIndustrial systems in the context of Industry 4.0 are facing ongoing challenges related to data consistency and interoperability, particularly in Digital Twin implementations. Although the Asset Administration Shell (AAS) provides a standardized foundation for structuring and exchanging industrial data, many systems still rely on custom formats, lim- iting reuse and integration. Creating and updating AAS repositories is time-consuming and requires specialized knowledge, and ensuring interoperability while maintaining flexibility and reusability remains a key challenge. To address these challenges, this dis- sertation proposes the development of a standardized Pose Submodel Template, designed according to the Industrial Digital Twin Association (IDTA) guidelines, to represent robotic poses in a structured and standard-compliant way. In parallel, a multi-agent system pow- ered by Large Language Models (LLMs) was implemented to automate AAS submodel generation, along with a Query–Response subsystem enabling natural language interac- tion with the existing AAS infrastructure, NOVAAS. The results show that combining standardized AAS submodels with generative AI effectively automates submodel creation, reduces human intervention, and facilitates data access, contributing to more scalable, interoperable, and intelligent Industry 4.0 systems.